Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add thaolst/ai-growth-prompts --skill 07-experiment-designgit clone --depth 1 https://github.com/thaolst/ai-growth-promptsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/thaolst/ai-growth-prompts/07-experiment-design)<a href="https://agentmods.dev/skills/thaolst/ai-growth-prompts/07-experiment-design"><img src="https://agentmods.dev/badge/skills/thaolst/ai-growth-prompts/07-experiment-design/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/thaolst/ai-growth-prompts/07-experiment-design"><img src="https://agentmods.dev/badge/skills/thaolst/ai-growth-prompts/07-experiment-design.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.03141 |
| Opus 5 | $0.00000 | $0.01571 |
| Sonnet 5 | $0.00000 | $0.00628 |
| Haiku 4.5 | $0.00000 | $0.00314 |
Grade A, and why
07-experiment-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 323 lines — stays where its author put it; the contents beside it link to each section on GitHub.
07 · Experiment design & A/B testing
Experiment là cách duy nhất để biết cái gì thực sự hiệu quả. Ở cấp S, experiment đơn giản: A/B 1 biến, đo 1 metric. Ở cấp M, có thể chạy multivariate, đo nhiều metric, có statistical significance. Ở cấp L, experiment là hệ thống: always-on testing, automated decision.
| Cấp độ | Experiment khả thi | Ràng buộc |
|---|---|---|
| S | A/B 1 biến, 1 metric, manual analyze | Traffic thấp, chạy 1-2 tuần, significance khó đạt |
| M | A/B đa biến, funnel analysis, cohort compare | Traffic vừa, có track tool, chạy 2-4 tuần |
| L | Always-on testing, automated decision, personalization | Full data pipeline, ML model, cross-team |
Prompt 15 · Thiết kế A/B test cho campaign
Khi nào dùng: Sắp launch campaign và muốn test trước 1 biến để tối ưu.
Cần thiết kế A/B test cho campaign sắp chạy.
Campaign: [mô tả]
Kênh: [in-app push / email / SMS / paid social]
Cấp độ: [S / M / L]
Traffic dự kiến: [số user / ngày]
Thời gian chạy: [số ngày]
Biến muốn test:
- [ ] Offer / voucher value
- [ ] Message / creative
- [ ] Timing / frequency
- [ ] Segment / targeting
- [ ] CTA / button
Yêu cầu output:
1. Hypothesis: If [biến] then [kết quả] because [lý do]
2. Control vs Treatment mô tả cụ thể
3. Primary metric (1 cái duy nhất)
4. Sample size cần: tính dựa trên baseline + MDE
5. Duration: bao nhiêu ngày để đủ significance
6. Risk: nếu test sai hoặc không đủ traffic thì sao
7. Decision rule: khi nào dừng test, khi nào chọn winner
Với [S]: chấp nhận directional signal (không cần p-value)
Với [M]: yêu cầu p < 0.05, ít nhất 80% power
Ví dụ output:
Hypothesis: If we change notification copy from "Ưu đãi hôm nay"
to "Chỉ còn 4 tiếng" then CTR increases by 15% because urgency
framing activates loss aversion.
Control: "Ưu đãi hôm nay — mở xem ngay"
Treatment: "Chỉ còn 4 tiếng — voucher của bạn sắp hết"
Primary metric: notification CTR
Sample size: 12,000 per group (based on 8% baseline CTR, MDE 15%, 95% confidence)
Duration: 10 days
Decision rule: stop if treatment CTR > control by ≥15% for 3 consecutive days
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 323 lines · 0 tokens per session scan A e742935f1e74
07-experiment-design is a skill published in the GitHub repository thaolst/ai-growth-prompts (11 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,141 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
multi-agent-workflow
Run a two-stage research then planning workflow for growth campaigns. Use when the user wants to analyze past campaign data AND create a new plan based on that analysis in one connected workflow. Stage 1: analyze data and extract structured insights. Stage 2: use insights to build a campaign plan. Input: past campaign…
ab-test-analyzer
Analyze A/B test results and recommend next actions. Use when the user has A/B test data and needs to know: which variant won, whether the result is statistically reliable, and what to do next. Works for conversion rate tests, copy tests, mechanic tests, UI tests. Input: user counts and conversions for control and…
campaign-brief
Write a complete campaign brief for growth marketing campaigns. Use when the user needs to create a campaign brief, plan a promotion, design a voucher campaign, or structure a marketing campaign. Works for fintech, e-commerce, super apps, and mobile payment platforms. Input: target segment, objective, budget, mechanic…
campaign-planning
Build a monthly or quarterly campaign plan from a growth target. Use when the user has a target (MEU, transactions, revenue, retention) and needs a full campaign plan: how many campaigns, what type, in what order, with what budget. Higher-level than campaign-brief (which covers a single campaign in detail). Input…
churn-intervention
Design churn prevention and winback campaigns for fintech and super apps. Use when the user needs to reduce churn, design save offers, create cancellation flows, plan re-engagement campaigns, or analyze churn reasons. Supports both proactive (predictive) and reactive (post-churn) strategies.
fintech-campaign-designer
Design full fintech campaigns from concept to execution — promotion strategy, mechanic, user journey, budget, and risk assessment. Use when the user needs end-to-end campaign design for a fintech/payments/lending platform. Specialized for SEA markets (Vietnam, Indonesia, Philippines, Thailand).